| Popis: |
Pri daljšem zajemanju slik z digitalnimi zrcalnorefleksnimi fotoaparati pogosto naletimo na težavo pregrevanja tipala fotoaparata, ki se izraža kot šum in druge nepravilnosti v digitalni sliki. Rešitev je zajem več slik, ki se zaporedno zajemajo manj časa, nato pa se zlijejo v eno skupno sliko. V tem diplomskem delu smo se ukvarjali s poravnavo in združevanjem (zlivanjem) več digitalnih astronomskih slik. Za združevanje slik smo uporabili postopek valčne transformacije ali metodo svetlejšega piksla, postopek poravnave digitalnih slik pa smo izvedli z oglišči, detektiranimi s Harrisovim detektorjem, in homografijo, ocenjeno z metodo RANSAC. Rezultat diplomskega dela je programsko orodje za zlivanje astronomskih digitalnih slik – AstrophotoTool, v katerega smo integrirali preučene algoritme za zlivanje in poravnavo slik. Za validiranje rezultatov smo uporabili lastno slikovno gradivo, ki smo ga zajeli na treh različnih lokacijah. Dobljene rezultate smo primerjali še z rezultati štirih uveljavljenih programskih orodij. Na osnovi analize rezultatov smo ugotovili, da z razvitim orodjem AstrophotoTool dobimo povsem primerne in pričakovane kvalitativne rezultate. V okviru tega diplomskega dela smo dobljene rezultate tudi kvantitativno ovrednotili. Iz slik, dobljenih po zlivanju, smo določili povprečno razmerje SNR in standardni odklon razmerij SNR, izračunanih v oknih velikosti 7x7 pikslov. Ugotovili smo, da se izračunana razmerja SNR za našo aplikacijo AstrophotoTool ujemajo z izračuni primerjanih aplikacij. S tem smo eksperimentalno potrdili pravilnost delovanja naše aplikacije AstrophotoTool. When we shoot long time exposures with digital SLR cameras, we very often face sensor overheating, which is visible on photos like noise or some other abnormalities. The best solution is to take multiple pictures one after another, by shooting each photo for small amount of time. Those pictures merged together give the same result as one picture taken for a longer time. For merging photos we used vawelet transformation or brighest pixel method. For registering digital images we used corners detected by Harris corner detector and homography estimated by RANSAC method. The result of this work represents a developed tool for merging astronomic digital images – AstrophotoTool. This tool includes all examined algorithms for merging and registration of the images. For validation we used our own photo material from three different locations. In the end we compared our results with results of four existing applications. From the analysis of the results (comparison of merged images of different applications), we found that developed tool AstrophotoTool gives expected and satisfying results. Obtained results were also quantitatively evaluated. We calculated the average SNR and standard deviations of SNRs. Those values were calculated from windows sized by 7x7 pixels. Conclusion is that SNR ratios for our application AstrophotoTool matches the calculation of compared applications. With that procedure we confirmed that our tool is working properly. |
